A method and system for monitoring and processing abnormal electrocardiogram acquisition points

Abstract

This application relates to the technical field of electrocardiogram (ECG) monitoring, and in particular to a method and system for monitoring and processing abnormal ECG acquisition points. The method includes: generating corresponding acquisition time points and reading time points in each sampling cycle based on action time setting rules, with a corresponding readback interval between the acquisition time points and reading time points; acquiring ECG data in each sampling cycle and storing it in a register after digital-to-analog conversion; triggering an acquisition task based on the acquisition time point to sample ECG data in the register and define it as an acquisition point; triggering a reading task based on the reading time point to read back the corresponding ECG data in the register and define it as a readback point; determining whether the data content of the acquisition point and the readback point in the same sampling cycle are the same, and filtering and processing abnormal points based on the determination result. This application effectively reduces the impact of abnormal ECG data identification caused by conflicts between Bluetooth tasks and acquisition tasks.

Description

Technical Field

[0001] This application relates to the technical field of electrocardiogram (ECG) monitoring, and in particular to a method and system for monitoring and processing abnormal ECG acquisition points. Background Technology

[0002] Dynamic electrocardiogram recorders have the functions of pacing detection and wireless transmission. The implementation of pacing detection requires sampling of electrocardiogram signals, and the sampling frequency usually cannot be lower than a certain threshold. Wireless transmission uses the Bluetooth protocol, which has very high timing requirements. During the execution of the application in the main control unit, the Bluetooth protocol stack will periodically interrupt the current task, jump to execute the Bluetooth task, and then jump back after completion. Since the execution time of the Bluetooth task is very short, it will not have much impact on general application scenarios.

[0003] However, when a Bluetooth SOC collects ECG signals, it needs to perform an ECG acquisition task every certain period of time, and the acquisition action must be completed within a specified time. If the Bluetooth task and the acquisition task conflict, the Bluetooth task will be executed first because it has a higher priority. When the Bluetooth task finishes and jumps back to the acquisition task, there is not enough time left to complete a normal acquisition action. This will eventually lead to abnormal ECG data. Abnormal data will cause the pacing detection algorithm to misidentify the data, which poses a significant safety hazard. Summary of the Invention

[0004] To reduce the impact of abnormal ECG data identification caused by conflicts between Bluetooth tasks and acquisition tasks, this application provides a method and system for monitoring and processing abnormal ECG acquisition points.

[0005] Firstly, this application provides a method for monitoring and processing abnormal electrocardiogram (ECG) acquisition points, employing the following technical solution:

[0006] A method for monitoring and processing abnormal electrocardiogram (ECG) acquisition points includes the following steps:

[0007] Based on the action time setting rules, corresponding acquisition time points and reading time points are generated in each sampling period, and the acquisition time points and reading time points are separated by a corresponding readback interval time.

[0008] Electrocardiogram (ECG) data is acquired during each sampling period and stored in a register after digital-to-analog conversion.

[0009] The acquisition task is triggered based on the acquisition time point to acquire the ECG data in the register and define it as the acquisition point;

[0010] Based on the reading time point, a reading task is triggered to read back the ECG data in the corresponding register and define it as the readback point;

[0011] Determine whether the data content of the sampling point and the readback point in the same sampling period is the same, and filter out abnormal points and process them based on the determination result.

[0012] In some embodiments, the corresponding acquisition time point and reading time point are generated in each sampling period based on the action time setting rules, including the following steps:

[0013] The number of anomalies occurring within a preset time period and the timestamp corresponding to each anomaly are obtained. An interval setting strategy is generated based on the number and the timestamp. The interval setting strategy includes symmetrical, asymmetrical, and golden window strategies.

[0014] If the number of abnormal points is not greater than the first threshold, the interval setting strategy is the symmetrical type;

[0015] If the number of anomalies is greater than the first threshold and the timestamps are not duplicated, the interval setting strategy is asymmetric.

[0016] If the number of anomalies is greater than the first threshold and the timestamps are repetitive, the interval setting strategy is the golden window type.

[0017] In some embodiments, the method of obtaining the number of anomalies occurring within a preset time period and the timestamp corresponding to each anomaly, and generating an interval setting strategy based on the number and the timestamps, further includes the following steps:

[0018] In the symmetrical configuration, the acquisition time point and the reading time point are respectively placed on both sides symmetrical to the midpoint of the cycle, and the standard duration of the action is configured towards the midpoint of the cycle to obtain the acquisition segment and the readback segment. The time length between the acquisition segment and the readback segment is defined as the readback interval time.

[0019] In the asymmetric method, the acquisition time point and the reading time point are set on both sides of the midpoint of the cycle, reducing the time distance between the acquisition time point and the starting point of the sampling cycle, reducing the time distance between the reading time point and the midpoint of the cycle, and configuring the standard duration of the action towards the midpoint of the cycle to obtain the acquisition segment and the readback segment. The time length between the acquisition segment and the readback segment is defined as the readback interval time.

[0020] In the golden window format, the timestamp of the repeated anomaly point is defined as the conflict time. The sampling period is divided into golden windows that do not contain the conflict time. The acquisition segment and the readback segment are set in the golden window based on the time length and continuity of the golden window. The time length between the acquisition segment and the readback segment is defined as the readback interval time.

[0021] In some embodiments, determining whether the data content of the sampling point and the readback point in the same sampling period is the same, filtering out abnormal points based on the determination result, includes the following steps:

[0022] If the data content of the acquisition point and the readback point in the same sampling period is the same, then the data content is retained and defined as valid acquisition;

[0023] If the data content of the sampling point and the readback point in the same sampling period are different, the data content in that sampling period is defined as an anomaly and removed.

[0024] In some embodiments, determining whether the data content of the sampling point and the readback point in the same sampling period is the same, filtering out and processing outliers based on the determination result, further includes the following steps:

[0025] Determine whether the standard duration of the action is less than the readback time interval;

[0026] If so, the ECG data in the register is repeated during the readback time interval to generate a verification point;

[0027] When the data content of the sampling point and the readback point in the same sampling period are different, the data content corresponding to the verification point is compared with the sampling point and the readback point respectively;

[0028] If there are identical matches, the data with the same matches will be retained and defined as valid data collection.

[0029] If no matching data is found, the data content in that sampling period is defined as an outlier and removed.

[0030] In some embodiments, acquiring electrocardiogram (ECG) data during each sampling period and storing it in a register after digital-to-analog conversion includes the following steps:

[0031] The first buffer and the second buffer are set in the register;

[0032] Obtain the number of leads and assign an identity label to each lead;

[0033] Based on the standard lead sequence, the ECG data after digital-to-analog conversion is stored in the corresponding register address in the first buffer according to the identity tag;

[0034] Based on the reconstructed lead sequence, the ECG data after digital-to-analog conversion is stored in the corresponding register address of the second buffer according to the identity tag.

[0035] In the standard lead sequence and the reconstructed lead sequence, the identification labels corresponding to the leads in the first half and the second half are different.

[0036] In some embodiments, a data acquisition task and a data read task are triggered to determine whether the data content of the acquisition point and the readback point in the same sampling period are the same. Based on the determination result, outliers are filtered and processed, including the following steps:

[0037] The electrocardiogram data in the second buffer are collected at the specified acquisition time point to obtain a reconstructed sequence;

[0038] The electrocardiogram data in the first buffer is read at the specified reading time point to obtain a standard sequence;

[0039] During the readback interval, the order of the recombined sequences is adjusted based on the standard lead order to obtain a verification sequence corresponding to the standard sequence;

[0040] After the standard sequence is read, the verification sequence is compared with the standard sequence, and the outliers are screened based on the comparison results.

[0041] In some embodiments, outliers are filtered and processed based on the judgment results, including the following steps:

[0042] The failure rate of leads is determined based on the comparison of the data content. The failure rate is characterized as the ratio between the number of leads that failed the comparison and the total number of leads.

[0043] If the proportion of failed leads is greater than a preset value, then the abnormal points in the sampling period are removed.

[0044] If the proportion of failed leads is not greater than a preset value, then the ECG data segments corresponding to the failed leads in the abnormal points are reconstructed.

[0045] In some of these embodiments,

[0046] Secondly, this application provides an abnormal electrocardiogram acquisition point monitoring and processing system, which adopts the following technical solution:

[0047] An abnormal electrocardiogram acquisition point monitoring and processing system is provided to implement the above method.

[0048] The technical solutions provided by the embodiments of this application can achieve the following technical effects:

[0049] The Bluetooth SOC performs two reading actions in each reading cycle and compares the two ECG data. If the two are completely consistent, the reading is considered successful; otherwise, the reading is considered abnormal. This effectively identifies and eliminates "singularities" in the collected data, reduces ECG data monitoring abnormalities caused by abnormal ECG data acquisition, and reduces potential safety risks. Attached Figure Description

[0050] Figure 1 This is a schematic diagram of the steps of the abnormal ECG acquisition point monitoring and processing method provided in this embodiment.

[0051] Figure 2 This is a simplified schematic diagram of the sampling period corresponding to different action time setting rules in the embodiments of this application.

[0052] Figure 3 This is a simplified schematic diagram of the sampling period corresponding to the redundancy verification in the embodiments of this application.

[0053] Figure 4 This is a waveform diagram before anomaly processing in the embodiments of this application.

[0054] Figure 5 This is a waveform diagram after the abnormal points have been processed in the embodiments of this application. Detailed Implementation

[0055] To better understand the purpose, technical solutions, and advantages of this application, it has been described and illustrated below with reference to the accompanying drawings and embodiments. However, those skilled in the art should understand that this application can be implemented without these details. In some cases, to avoid obscuring various aspects of this application due to unnecessary description, well-known methods, processes, systems, components, and / or circuits already described at a higher level will not be elaborated upon. It will be apparent to those skilled in the art that various modifications can be made to the embodiments disclosed in this application, and the general principles defined in this application can be applied to other embodiments and application scenarios without departing from the principles and scope of this application. Therefore, this application is not limited to the illustrated embodiments, but conforms to the broadest scope consistent with the scope of protection claimed in this application.

[0056] It should be noted that the descriptions of these embodiments are for the purpose of aiding understanding the present invention, but do not constitute a limitation thereof. Furthermore, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0057] In the description of this application, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0058] In the description of this application, the terms "one embodiment," "some embodiments," "illustrative embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples.

[0059] like Figure 1 As shown in the figure, this application discloses a method for monitoring and processing abnormal electrocardiogram acquisition points, including the following steps:

[0060] S100 generates corresponding acquisition time points and reading time points in each sampling cycle based on the action time setting rules, with a corresponding readback interval time between the acquisition time points and reading time points.

[0061] This technology is applied to electrocardiogram (ECG) recorders, which need to periodically collect and analyze users' ECG data according to a certain acquisition cycle. In order to avoid the Bluetooth SOC being conflicted by other Bluetooth tasks with higher priority when it needs to collect ECG data, this application uses an ECG AFE chip to change the single reading action in the prior art into two readings. The first reading is for ECG data acquisition, and the second reading is for ECG data rereading. Since the ECG data read each time should be exactly the same, the consistency of the two acquired data can be used to determine whether there are abnormal reading points.

[0062] Therefore, it is necessary to first configure the acquisition time point corresponding to the acquisition action in each sampling period, and at the same time configure the reading time point corresponding to the second reading action.

[0063] At the same time, the Bluetooth SOC needs a certain response time for each read. Therefore, in order to avoid reading errors caused by insufficient response time between two read operations, a certain time interval needs to be set between the two reads, which is the readback interval.

[0064] Therefore, a sampling period consists of three time segments: the sampling action time segment starting from the sampling time point, the readback interval time, and the read time segment starting from the read time point.

[0065] S200 acquires ECG data in each sampling cycle and stores it in a register after digital-to-analog conversion.

[0066] The internal ADC of the ECG AFE chip samples the ECG signal at a frequency of 8kHz (i.e., completes an analog-to-digital conversion every 125μs), and the sampled ECG data is stored in a register. The sampled ECG data is a string of numbers of fixed byte length. At the same time, the ECG data of different leads corresponds to different numbers of bytes. For example, the ECG data of 12 leads is usually 24 bytes / sample, with 2 bytes of ADC value per lead.

[0067] After the conversion is complete, the Bluetooth SOC is notified to read the data via a hardware interrupt (such as the DRDY pin) or the SPI / I2C bus status.

[0068] After receiving the notification, the Bluetooth SOC will read the ECG data from the register at different times.

[0069] S300 triggers a data acquisition task based on the acquisition time point to acquire ECG data in the register and define it as the acquisition point.

[0070] When the acquisition time point arrives, the ECG AFE notifies the Bluetooth SOC to read the ECG data from the register, and then uses the data acquired during this time point as the acquisition point.

[0071] S400 triggers a read task based on the read time point to read back the ECG data in the corresponding register and defines it as the readback point.

[0072] Similarly, after waiting for the readback interval, the ECG AFE notifies the Bluetooth SOC to read the ECG data from the register again, and uses the data read back during this time interval as the readback point.

[0073] S500 determines whether the data content of the acquisition point and the readback point in the same sampling period are the same, and filters out abnormal points and processes them based on the judgment result.

[0074] If ECG data needs to be acquired every 125µs during a sampling cycle, the Bluetooth SOC must complete the entire acquisition process within 125µs. If a complete acquisition takes 40µs, and the Bluetooth SOC is occupied by other Bluetooth tasks, resulting in only 30µs of remaining acquisition time in that cycle when it resumes the acquisition process, then the acquired data will be abnormal.

[0075] In existing technologies where data is collected only once per sampling cycle, it is impossible to determine whether abnormal data is caused by a conflict between the Bluetooth task and the data collection task or by an abnormality in the user's health.

[0076] This application uses a method of reading ECG data twice and comparing the results. When it is known that two data can be read in one sampling cycle, it can know that there is no abnormality in the sampling cycle based on the fact that the comparison data content is the same. When the comparison data content is different, it can know that there is a Bluetooth SOC conflict sampling abnormality in the sampling cycle.

[0077] Through the above technical solution, the Bluetooth SOC performs two reading actions in each reading cycle and compares the ECG data content read twice. If the two are completely consistent, the reading is considered successful; otherwise, the reading is considered abnormal. This effectively identifies and eliminates "singularities" generated in the collected data, reduces ECG data monitoring abnormalities caused by abnormal ECG data acquisition, and reduces safety risks.

[0078] like Figure 2 As shown, in some other embodiments, the corresponding acquisition time point and reading time point are generated in each sampling period based on the action time setting rules, including the following steps:

[0079] S110: Obtain the number of anomalies occurring within a preset time period and the timestamp corresponding to each anomaly, and generate an interval setting strategy based on the number and timestamp. The interval setting strategy includes symmetrical, asymmetrical, and golden window.

[0080] The action time setting rule is a rule that optimizes the time interval between two reading actions in a cycle according to different situations. The start time of the reading action is adjusted according to the number of anomalies, their occurrence time, etc., in order to minimize the impact of Bluetooth task conflicts on ECG data acquisition when Bluetooth SOC occurs.

[0081] Since anomalies are characterized by conflicts between Bluetooth tasks and data acquisition tasks, the frequency, duration, and cycle of Bluetooth tasks can be analyzed by examining the number and timing of these anomalies.

[0082] The timestamp of the anomaly point represents the time when the anomaly occurred. Specifically, since each ECG data is proportional to the number of leads on the device (e.g., each lead corresponds to 2 bytes, so 12 leads correspond to 24 bytes), the time when the anomaly occurred can be generated based on the position of the first different byte after comparing two data readings. The time analyzed above may correspond to the time when the anomaly occurred during the first reading or the time when the anomaly occurred during the second reading. Therefore, it is only necessary to compare the difference between the two Bluetooth SOC readings and the ECG data initially acquired by the ECG AFE to determine which reading caused the anomaly.

[0083] In order to reduce abnormal ECG data monitoring caused by Bluetooth task conflicts, it is necessary to dynamically adjust the ECG data reading time based on the analyzed Bluetooth task characteristics to avoid the Bluetooth task execution time.

[0084] Interval setting strategies include symmetrical, asymmetrical, and golden window styles. Specifically,

[0085] S120, if the number of outliers is not greater than the first threshold, the interval setting strategy is symmetrical.

[0086] In the symmetrical configuration, the acquisition time point and the reading time point are placed on opposite sides of the midpoint of the cycle, and the standard duration of the action is configured towards the midpoint of the cycle to obtain the acquisition segment and the readback segment. The time length between the acquisition segment and the readback segment is defined as the readback interval time.

[0087] The standard duration of an action represents the time required for a complete reading task. The standard duration of an action will vary depending on the different acquisition precision and requirements.

[0088] When the number of anomalies is small, such as the extremely low probability of a conflict between the Bluetooth task and the acquisition task, with an average of one such conflict per second at an 8kHz sampling rate, then its impact on the overall ECG monitoring is low.

[0089] The symmetrical time interval setting method is also the conventional time interval setting method for Bluetooth readback. It is relatively simple to configure, and the frequency of ECG AFE notifying Bluetooth SOC to collect data is relatively uniform. It can be achieved by directly using a timer with a fixed time.

[0090] Specifically, in a 125us sampling period, sampling begins at 15us and continues until 55us. After another 7.5us, the period ends at the midpoint. After the same 7.5us, the reading time point is reached and reading begins (70us). The process continues until 110us, with the remaining 15us serving as response time and data comparison time.

[0091] S130, if the number of outliers is greater than the first threshold and there is no repetition among the timestamps, the interval setting strategy is asymmetric.

[0092] In the asymmetric mode, the acquisition time point and the reading time point are set on both sides of the midpoint of the cycle to reduce the time distance between the acquisition time point and the starting point of the sampling cycle, and to reduce the time distance between the reading time point and the midpoint of the cycle. The standard duration of the action is configured towards the midpoint of the cycle to obtain the acquisition segment and the readback segment. The time length between the acquisition segment and the readback segment is defined as the readback interval time.

[0093] When there are a large number of anomalies, and the timestamps corresponding to these anomalies are not repeated, it indicates that there have been a large number of relatively random Bluetooth task conflicts. In Bluetooth tasks, relatively random tasks include asynchronous events corresponding to data requests from external devices, such as commands sent by the app.

[0094] To reduce the probability of such conflicts occurring, it is necessary to properly manage the time interval between two reads.

[0095] In this application, by adjusting the conventional symmetrical dual-reading time setting to an asymmetrical one, the acquisition and readback times within a single sampling cycle are advanced. Although the priority of the Bluetooth task is higher than that of ECG data acquisition, by preempting task time slices, the read time within a single acquisition cycle is advanced as much as possible. Triggering an interrupt requires meeting hardware conditions (such as the Bluetooth chip sending an interrupt signal and the SOC's interrupt enable being enabled). If the read operation is completed before the interrupt signal arrives, conflicts can be minimized.

[0096] In other words, the difference between asymmetric and symmetric methods lies in the fact that the time of two reads is offset within a single sampling period to keep the non-reading time periods on the same side as much as possible, reducing the distribution of non-reading time periods. This makes it possible for the non-reading time periods that can be matched when relatively random Bluetooth task conflicts occur to be longer, thus reducing the probability of impacting the read time.

[0097] In other embodiments, the "tight at the beginning and loose at the end" approach in the above asymmetric method can be adjusted to a "loose at the beginning and tight at the end" approach based on the actual improvement in the number of anomalies, that is, the time of both readings is set synchronously in the direction of the later acquisition time.

[0098] S140. If the number of outliers is greater than the first threshold and the timestamps are repetitive, the interval setting strategy is the golden window.

[0099] In the golden window method, the timestamp of the repeated anomaly is defined as the conflict time. The sampling period is divided into golden windows that do not contain conflict times. Based on the time length and continuity of the golden window, the acquisition segment and the readback segment are set in the golden window. The time length between the acquisition segment and the readback segment is defined as the readback interval.

[0100] If a large number of outliers are collected, and analysis of the timestamps of several outliers reveals that the occurrence time periods of multiple outliers are repetitive, the repetition is characterized by the fact that the time positions of the outlier occurrence periods are the same or similar in each sampling period.

[0101] When repetitive events occur, it is assumed that the Bluetooth SOC performs some periodic events based on the protocol, such as connection intervals and heartbeat packet interactions. These types of data are triggered at fixed time intervals, and after being triggered, they will affect the collection of ECG data based on high priority.

[0102] In this case, the application proposes a golden window-style spacing based on the traditional symmetrical spacing and the improved asymmetrical spacing.

[0103] Because outliers are repetitive, the frequency of repetition can be used to analyze the position of the Bluetooth task conflict-prone time of the Bluetooth SOC in each sampling period. The golden window is characterized by avoiding the Bluetooth task conflict-prone time between two readings, and the readback interval between two readings is used as the golden window to match the Bluetooth task conflict-prone time. In this way, the occurrence of conflicts can be reduced by dynamically adjusting the reading time.

[0104] If, after analyzing the timestamps of outliers, it is found that outliers frequently occur between 50us and 65us in the sampling period, then a golden window is set within the 50us-65us range, and the two reading times need to avoid the golden window. In this way, the reading time position can be dynamically adjusted according to the periodicity of the outlier occurrence time to reduce the occurrence of conflicts.

[0105] It is important to note that since the time required for each read operation must meet certain requirements, if the remaining time period after setting the golden window in the sampling period does not meet the time length required for two complete reads, then the golden window interval setting strategy should be prohibited in that sampling period.

[0106] In other embodiments, it is determined whether the data content of the acquisition point and the readback point in the same sampling period are the same, and anomalies are filtered and processed based on the determination result, including the following steps:

[0107] S510: If the data content of the acquisition point and the readback point in the same sampling period is the same, then the data content is retained and defined as valid acquisition.

[0108] If the data content of the acquisition point and the readback point in a sampling period are completely consistent, it is considered that the Bluetooth SOC did not read less or misread the ECG data in the register due to the conflict between the Bluetooth task and the acquisition task in that time period, and the data content is defined as valid acquisition.

[0109] S520: If the data content of the acquisition point and the readback point in the same sampling period are different, the data content in that sampling period is defined as an outlier and removed.

[0110] If the data content of the acquisition point and the readback point in a sampling period are inconsistent, it is considered that the Bluetooth SOC did not perform the reading task when reading data once or twice due to the conflict between the Bluetooth task and the acquisition task, but instead processed the higher priority Bluetooth task. Therefore, this data content is defined as an abnormal point and is removed.

[0111] Among these, the abnormal situations caused by conflicts include several types:

[0112] Both reads were incomplete due to conflicts.

[0113] One reading was completed, but another reading was incomplete due to a conflict, only reading ECG data for a period of time after the Bluetooth SOC finished processing the Bluetooth task and restored the data acquisition task.

[0114] One read operation is completed, but another read operation results in a conflict that causes only a portion of the ECG data for the current sampling period to be read. When the Bluetooth SOC continues to read, the time reaches the next sampling period, and the ECG AFE automatically acquires the ECG data for the next period and places it in the register after digital-to-analog conversion. As a result, the Bluetooth SOC reads a portion of the ECG data from the previous sampling period and a portion of the ECG data from the next sampling period within the time of one read task.

[0115] like Figure 3 As shown, in some other embodiments, determining whether the data content of the sampling points and readback points in the same sampling period is the same, filtering out abnormal points based on the determination result, and processing them further includes the following steps:

[0116] S530, determine whether the standard duration of the action is less than the readback time interval.

[0117] Under normal circumstances, the probability of a conflict between Bluetooth tasks and data acquisition tasks is relatively low. At a sampling rate of 8kHz, an average of one conflict will occur per second. If the abnormal points are directly removed, the impact on the overall sampling data and pacing detection is negligible.

[0118] However, in situations requiring high sampling accuracy or with heavy Bluetooth workloads, the number of collisions per second can be relatively high. Therefore, to address the high accuracy requirements of ECG acquisition, this application further incorporates redundant decision-making steps. The overall logic optimizes the two-step reading decision-making process within each sampling period to three-step reading decision-making processes, provided the sampling period length allows.

[0119] Based on redundancy determination, it is first necessary to determine whether the time interval between or outside two reads allows the third read action to be completed. Therefore, it is necessary to determine whether the standard duration of the action is less than the read interval.

[0120] S531, if so, then the ECG data in the register is repeated during the readback interval to generate a verification point.

[0121] If the time is less than the required time, it means that there is enough time remaining to perform one readback action. In this case, the ECG AFE notifies the Bluetooth SOC to perform another readback action to read the ECG data in the register and define it as a verification point.

[0122] Verification points are used to perform redundancy verification on ECG data from two regular reads.

[0123] S532, when the data content of the acquisition point and the readback point in the same sampling period are different, the data content corresponding to the verification point is compared with the acquisition point and the readback point respectively.

[0124] When the data content of the acquisition point and the readback point in the same sampling period are different, the scheme based on the above embodiment will directly remove the data in that sampling period. So, after adding redundancy verification, instead of removing the abnormal points, the verification points are compared with the acquisition points and the readback points respectively, and the subsequent processing method is selected based on the comparison results.

[0125] S533: If there are identical data points, retain the identical data points and define them as valid acquisitions.

[0126] S534 If no matching data is found, the data content in that sampling period is defined as an outlier and removed.

[0127] The above describes several situations where conflicts between Bluetooth tasks and acquisition tasks lead to abnormal ECG data acquisition.

[0128] When two readings are incomplete due to conflicts, the data from the two readings will be different and will not match the data at the verification point. Therefore, the ECG data in that sampling period will be directly removed.

[0129] If one read is complete while another is incomplete, two scenarios will occur. If the ECG data at the verification point is collected completely without conflict, the ECG data at the verification point will match the ECG data at the acquisition point or readback point. This indicates that at least one ECG data sampling was complete and correct, and the ECG data that matches perfectly can be retained as valid acquisition. However, if the ECG data at the verification point is also incomplete due to Bluetooth conflict, even if at least one of the acquisition point or readback point has correct data, it cannot match the ECG data at the verification point. In this case, the system cannot determine which data reading is valid, and the ECG data in that sampling period will be directly discarded as an anomaly.

[0130] By using the above method, the probability and accuracy of correctly reading and judging ECG data can be improved by repeatedly comparing the redundantly set verification points before Bluetooth conflicts occur or after conflicts end. To a certain extent, this preserves ECG data that could not be effectively captured due to conflicts, reduces the number of abnormal points, and makes the ECG monitoring data more complete and consistent.

[0131] In the above solutions, all methods involve storing the ECG data sampled by the AFE in a register, while the Bluetooth SOC reads the ECG data from a single register during a sampling cycle. In other embodiments, a partitioned storage and partitioned reading approach can be used, combining the actions and results of the two readings. Specifically, ECG data is acquired in each sampling cycle, converted from digital to analog, and then stored in a register, including the following steps:

[0132] S210 sets the first and second buffers in the register.

[0133] A buffer is defined as a region set up inside a register for temporary data storage. Buffers include data buffers (DR), address buffers (AR), instruction buffers (IR), and I / O buffers. The buffer mentioned in this application is a data buffer.

[0134] There are two buffers. After the ECG AFE samples ECG data, it will store the sampled ECG data in the first buffer and the second buffer respectively according to different rules.

[0135] S220: Obtain the number of leads and set an identity label for each lead.

[0136] Each lead has a corresponding identification label based on its number and location. For example, lead I represents the limb lead and corresponds to the lateral wall of the heart, lead II represents the limb lead and corresponds to the anterior wall of the heart, and lead V1 is located at the right sternal border in the fourth intercostal space and represents the anterior septal wall of the heart.

[0137] Different numbers of leads correspond to different identification labels, and these labels can be modified and formulated based on actual conditions and domestic and international requirements.

[0138] S230, based on the standard lead sequence, stores the ECG data after digital-to-analog conversion in the corresponding register address in the first buffer according to the identity tag.

[0139] The standard lead sequence is characterized by the standard order in which ECG data from multiple leads are sampled by an AFE (Automatic External Defibrillator) and stored in registers after digital-to-analog processing. For example, the typical output sequence for a 12-lead ECG is:

[0140] Lead 1 (I) - Lead 2 (II) - Lead 3 (III) - Lead 4 (avR) - Lead 5 (aVL) - Lead 6 (aVF) - Lead 7 (V1) - ... - Lead 12 (V6).

[0141] The ECG data sampled from several leads are stored sequentially in the register addresses of the first buffer in the above order. The storage structure of the first buffer is then as follows:

[0142] EG_ADDR[0x00] = Low byte of lead 1;

[0143] REG_ADDR[0x01] = High byte of lead 1;

[0144] REG_ADDR[0x02] = Low byte of lead 2;

[0145] REG_ADDR[0x03] = High byte of lead 2; ...

[0146] REG_ADDR[0x16] = Lead 12 high byte.

[0147] S240, based on the reorganized lead sequence, stores the ECG data after digital-to-analog conversion in the corresponding register address in the second buffer according to the identity tag.

[0148] The reconfigured lead sequence is a new sorting rule generated after adjusting the position of the standard lead sequence. At the same time, the standard lead sequence and the reconfigured lead sequence need to meet the following condition: in the standard lead sequence and the reconfigured lead sequence, the identification labels corresponding to the leads in the first half and the second half are different.

[0149] In other words, the partitions of each lead in the standard lead sequence and the reconstructed lead sequence are different. In one case, the reconstructed lead sequence of 12 leads can be:

[0150] Lead 7 (V1) - ... - Lead 12 (V6) - Lead 1 (I) - Lead 2 (II) - Lead 3 (III) - Lead 4 (avR) - Lead 5 (aVL) - Lead 6 (aVF).

[0151] This allows the leads in each position in both order rules to be in different positions without completely scrambling the lead positions.

[0152] The lead order can be rearranged and remapped using array indexing.

[0153] In other embodiments, a data acquisition task and a data read task are triggered to determine whether the data content of the acquisition points and readback points in the same sampling period are the same. Based on the determination result, outliers are filtered and processed, including the following steps:

[0154] S540 samples the electrocardiogram data in the second buffer at the acquisition time point to obtain the recombinant sequence.

[0155] When setting up the buffer, the objects corresponding to the first and second read actions are different. In the first read, the ECG AFE first notifies the Bluetooth SOC at the acquisition time point to read the ECG data in the second buffer to obtain the recombined sequence based on the recombined lead order.

[0156] S541, read the ECG data in the first buffer at the read time point to obtain the standard sequence.

[0157] During the second reading, the ECG AFE notifies the Bluetooth SOC at the reading time point to read the ECG data in the first buffer to obtain a standard sequence based on the standard lead order.

[0158] S542, during the readback interval, adjusts the order of the recombined sequence based on the standard lead sequence to obtain a verification sequence corresponding to the standard sequence.

[0159] The only difference between the recombinant sequence and the standard sequence is the lead order; the number of bytes and the byte content of each lead should remain consistent unless there are any conflicts or anomalies.

[0160] Reordering multi-lead data may cause confusion in the physical location and electrical characteristics of leads, resulting in waveform distortion. The purpose of adjusting the lead order in this application is to ensure that the reading order of each lead is different in each reading process, thereby reducing the probability of frequent Bluetooth conflicts affecting a certain part of the ECG data. Therefore, in order to avoid morphological distortion, the order of the reconstructed sequence needs to be adjusted to restore it to the same order as the standard sequence when comparing two readings.

[0161] After the first read is completed and before the second read begins, the Bluetooth SOC rearranges the lead order in memory to make the recombined sequence consistent with the standard sequence and redefines it as the verification sequence.

[0162] S543, after the standard sequence is read, the verification sequence is compared with the standard sequence, and outliers are screened based on the comparison results.

[0163] The verification sequence is compared with the standard sequence, and if there are inconsistent byte contents during the comparison, the ECG data in that sampling period is regarded as an outlier.

[0164] By setting up buffers, multiple read tasks from a single storage location are transformed into multiple read tasks from multiple storage locations with different timing sequences, making the anomaly analysis process more accurate.

[0165] In other embodiments, outliers are filtered and processed based on the judgment results, including the following steps:

[0166] S550 determines the percentage of leads that failed to match based on the comparison of data content. The percentage of leads that failed to match is represented as the ratio between the number of leads that failed to match and the total number of leads.

[0167] Furthermore, after the outliers are identified, in addition to retaining or removing them through redundancy checks, the ECG data corresponding to the outliers can also be reconstructed using existing lead reconstruction methods.

[0168] However, on the one hand, the reconstructed lead data has a certain predictive power, and on the other hand, the confidence level of the reconstructed data varies under different scenarios. Therefore, in order to further improve the accuracy and appropriateness of outlier data processing, this application further determines the final outlier processing method based on the proportion of failed leads.

[0169] First, the number of mismatched leads is determined by comparing the consistency of the data content, and then the percentage of failed leads is calculated based on the total number of leads.

[0170] S551: If the proportion of failed leads is greater than a preset value, then the abnormal points in this sampling period will be removed.

[0171] If the proportion of failed leads is large, it means that a large number of leads were not read correctly due to conflicts. Since lead reconstruction also needs to be based on the data of other related leads in the same sampling period and the data of the lead in other sampling periods, when the erroneous lead data accounts for a large proportion, the confidence of the reconstructed data is low and may not meet the needs of clinical medicine. Therefore, the abnormal point can be directly removed.

[0172] S552, if the proportion of failed leads is not greater than a preset value, then the ECG data segments corresponding to the failed leads in the abnormal points are reconstructed.

[0173] If the proportion of failed leads is small, the confidence level of the reconstructed lead data will be high because there are many other related correct lead data at the same time. In this case, the missing lead data in the abnormal points can be reconstructed based on the existing ECG data reconstruction methods.

[0174] Refactoring methods include:

[0175] Within the same sampling period, by utilizing the spatial electrical conduction correlation between multiple leads, the missing lead is predicted through linear or nonlinear combinations of known leads. Specifically:

[0176] Linear transformation method: directly calculate the lost lead through linear combination of other leads (e.g., III=II-I).

[0177] Graph Neural Networks (GNNs): Treat leads as nodes in a graph structure and learn spatial dependencies between leads (such as the proximity relationship between V1 and V2) through graph convolution. They are suitable for scenarios where multiple chest leads are lost.

[0178] An autoencoder learns the overall features of 12 leads through a neural network. When the input contains missing leads, the model reconstructs the missing values ​​by associating features from other leads.

[0179] Based on historical data from lost leads, the temporal continuity analysis of ECG signals (P wave, QRS wave, T wave) is used to predict missing segments from the historical waveforms of the lost leads. Specifically:

[0180] Time-domain interpolation methods: cubic spline interpolation and piecewise linear interpolation are suitable for short time gaps, directly filling intermediate values ​​based on the waveform shape of the preceding and following time points;

[0181] Kalman filtering treats the ECG signal as a dynamic system, using historical states and current observations to predict the temporal changes of missing leads, and is suitable for handling short-term missing leads with noise.

[0182] Recurrent Neural Networks (RNN / LSTM): By learning the historical temporal characteristics of missing leads (such as QRS frequency and ST segment morphology), they can predict the waveform of future missing segments.

[0183] The improvement effect on ECG data achieved through the analysis and processing of outliers in this application is as follows: Figure 4 , Figure 5 As shown, the abnormal points in the final electrocardiogram waveform were processed, and the electrocardiogram data tended to be normal.

[0184] It also includes the following steps:

[0185] The S600 determines health scenarios based on effectively acquired electrocardiogram (ECG) data.

[0186] Based on the effectively collected ECG data, a corresponding ECG monitoring waveform is generated. The waveform is then used to analyze whether there are any health abnormalities. For example, stable and regular ECG data corresponds to a normal healthy scenario; if the ECG data is unstable and has many high and low points, it corresponds to an unhealthy scenario.

[0187] S610 generates corresponding risk ratings based on health scenarios.

[0188] The sensitivity to risk assessment of ECG data monitoring varies depending on the specific ECG monitoring scenario. The occurrence of anomalies can lead to the loss of some ECG data, and the loss of ECG data corresponds to different risk ratings in different scenarios.

[0189] A higher risk rating indicates a greater impact from data loss in that scenario, while a lower risk rating indicates a smaller impact from data loss in that scenario.

[0190] S620 marks anomalies based on risk rating and generates alarm information based on the timestamps corresponding to the anomalies.

[0191] Different risk ratings correspond to different anomaly alert strategies. When anomalies occur and are processed such as removal, reconstruction, or redundant judgment retention in low-risk and high-risk situations, the anomaly needs to be marked regardless of whether it is retained. This is to inform the user that the point is not a data point collected normally, but rather the result of the processing of the anomaly, which the user needs to view and confirm.

[0192] Meanwhile, in low-risk situations, since users' health data is relatively stable, alarm thresholds for anomalies can be set. For example, an alarm will only be issued if the frequency of anomalies exceeds a preset value within a certain period of time or if anomalies occur several times consecutively within a fixed time. In high-risk situations, since users' overall health data shows an unfavorable risk, an alarm can be issued every time an anomaly occurs to prompt users or administrators to check and confirm.

[0193] Under different risk levels, the sound and light alerts for anomaly point alarms can be different. Because the anomaly point alarm under low risk is mainly used to indicate that the frequency and number of anomalies are high, reminding managers to make corresponding task adjustments and optimizations to the Bluetooth SOC, while the anomaly point alarm under high risk is mainly used to remind that when anomalies occur, some risky ECG data may be ignored or mishandled, leading to safety hazards. Therefore, different photoelectric alarm methods are used to classify and distinguish them.

[0194] The implementation principle is as follows:

[0195] The Bluetooth SOC performs two reading actions in each reading cycle and compares the two ECG data. If the two are completely consistent, the reading is considered successful; otherwise, the reading is considered abnormal. This effectively identifies and eliminates "singularities" in the collected data, reduces ECG data monitoring abnormalities caused by abnormal ECG data acquisition, and reduces potential safety risks.

[0196] It should be understood that although the steps in the flowcharts in the accompanying drawings are shown sequentially as indicated by the arrows, these steps are not necessarily performed in the order indicated by the arrows. Unless otherwise expressly stated herein, there is no strict order in which these steps are performed, and they may be performed in other orders.

[0197] The above are all preferred embodiments of the present application, and are not intended to limit the scope of protection of the present application. Therefore, any equivalent changes made based on the structure, shape, and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A method for monitoring and processing abnormal electrocardiogram (ECG) acquisition points, characterized in that, Includes the following steps: Based on the action time setting rules, corresponding acquisition time points and reading time points are generated in each sampling period, and the acquisition time points and reading time points are separated by a corresponding readback interval time. Electrocardiogram (ECG) data is acquired during each sampling period and stored in a register after digital-to-analog conversion. The acquisition task is triggered based on the acquisition time point to acquire the ECG data in the register and define it as the acquisition point; Based on the reading time point, a reading task is triggered to read back the ECG data in the corresponding register and define it as the readback point; Determine whether the data content of the collection point and the readback point in the same sampling period is the same, filter out abnormal points based on the judgment result and process them, the processing includes elimination and retention; The method for generating the action time setting rules specifically includes: The number of anomalies occurring within a preset time period and the timestamp corresponding to each anomaly are obtained. An interval setting strategy is generated based on the number and the timestamp. The interval setting strategy includes symmetrical, asymmetrical, and golden window strategies. If the number of abnormal points is not greater than the first threshold, the interval setting strategy is the symmetrical type; If the number of anomalies is greater than the first threshold and the timestamps are not duplicated, the interval setting strategy is asymmetric. If the number of anomalies is greater than the first threshold and the timestamps are repetitive, the interval setting strategy is a golden window. Specifically, in the golden window, the timestamps of the repeated anomalies are defined as conflict times. The sampling period is divided into golden windows that do not contain the conflict times. Based on the time length and continuity of the golden window, the acquisition segment and the readback segment are set in the golden window. The time length between the acquisition segment and the readback segment is defined as the readback interval time.

2. The method for monitoring and processing abnormal ECG acquisition points according to claim 1, characterized in that: The method includes obtaining the number of anomalies occurring within a preset time period and the timestamp corresponding to each anomaly, and generating an interval setting strategy based on the number and the timestamps. The method further includes the following steps: In the symmetrical configuration, the acquisition time point and the reading time point are respectively placed on both sides symmetrical to the midpoint of the cycle, and the standard duration of the action is configured towards the midpoint of the cycle to obtain the acquisition segment and the readback segment. The time length between the acquisition segment and the readback segment is defined as the readback interval time, and the standard duration of the action represents the time required for a complete reading task. In the asymmetric method, the acquisition time point and the reading time point are set on both sides of the midpoint of the cycle, reducing the time distance between the acquisition time point and the starting point of the sampling cycle, reducing the time distance between the reading time point and the midpoint of the cycle, and configuring the standard duration of the action towards the midpoint of the cycle to obtain the acquisition segment and the readback segment. The time length between the acquisition segment and the readback segment is defined as the readback interval time.

3. The method for monitoring and processing abnormal ECG acquisition points according to claim 1, characterized in that, Determine whether the data content of the sampling point and the readback point in the same sampling period are the same, filter out abnormal points based on the determination result, and process them, including the following steps: If the data content of the acquisition point and the readback point in the same sampling period is the same, then the data content is retained and defined as valid acquisition; If the data content of the sampling point and the readback point in the same sampling period are different, the data content in that sampling period is defined as an anomaly and removed.

4. The method for monitoring and processing abnormal ECG acquisition points according to claim 2, characterized in that, Determining whether the data content of the sampling point and the readback point in the same sampling period is the same, filtering out and processing outliers based on the determination result, also includes the following steps: Determine whether the standard duration of the action is less than the readback interval time; If so, the ECG data in the register is repeated during the readback interval to generate a verification point; When the data content of the sampling point and the readback point in the same sampling period are different, the data content corresponding to the verification point is compared with the sampling point and the readback point respectively; If there are identical matches, the data with the same matches will be retained and defined as valid data collection. If no matching data is found, the data content in that sampling period is defined as an outlier and removed.

5. The method for monitoring and processing abnormal ECG acquisition points according to claim 1, characterized in that, Acquiring ECG data during each sampling period and storing it in a register after digital-to-analog conversion includes the following steps: The first buffer and the second buffer are set in the register; Obtain the number of leads and assign an identity label to each lead; Based on the standard lead sequence, the ECG data after digital-to-analog conversion is stored in the corresponding register address in the first buffer according to the identity tag; Based on the reconstructed lead sequence, the ECG data after digital-to-analog conversion is stored in the corresponding register address of the second buffer according to the identity tag. In the standard lead sequence and the reconstructed lead sequence, the identification labels corresponding to the leads in the first half and the second half are different.

6. The method for monitoring and processing abnormal ECG acquisition points according to claim 5, characterized in that, Triggering a data acquisition task and a data readout task, determining whether the data content of the acquisition point and the readback point in the same sampling period are the same, filtering out abnormal points based on the determination result and processing them, including the following steps: The electrocardiogram data in the second buffer are collected at the specified acquisition time point to obtain a reconstructed sequence; The electrocardiogram data in the first buffer is read at the specified reading time point to obtain a standard sequence; During the readback interval, the order of the recombined sequences is adjusted based on the standard lead order to obtain a verification sequence corresponding to the standard sequence; After the standard sequence is read, the verification sequence is compared with the standard sequence, and the outliers are screened based on the comparison results.

7. The method for monitoring and processing abnormal ECG acquisition points according to claim 1, characterized in that, Based on the judgment results, outliers are filtered and processed, including the following steps: The failure rate of leads is determined based on the comparison of the data content. The failure rate is characterized as the ratio between the number of leads that failed the comparison and the total number of leads. If the proportion of failed leads is greater than a preset value, then the abnormal points in the sampling period are removed. If the proportion of failed leads is not greater than a preset value, then the ECG data segments corresponding to the failed leads in the abnormal points are reconstructed.

8. The method for monitoring and processing abnormal electrocardiogram acquisition points according to claim 3 or 4, characterized in that, It also includes the following steps: Based on the effectively collected electrocardiogram (ECG) data, a health scenario is determined, which includes daily health scenarios where the ECG data is stable and regular, and unhealthy scenarios where the ECG data is unstable. A corresponding risk rating is generated based on the health scenario. The higher the risk rating, the greater the impact of data loss in that scenario. The lower the risk rating, the smaller the impact of data loss in that scenario. Here, the lost data corresponds to the occurrence of the anomaly. The anomalies are marked based on the risk rating, and alarm information is generated based on the timestamps corresponding to the anomalies.

9. A monitoring and processing system for abnormal electrocardiogram acquisition points, characterized in that, Used to implement the method as described in any one of claims 1-8.

Images